The global financial crisis brought with it fear. No small amount of it was focused on the possibility of extreme 1930s-style protectionism. Thanks to an ambitious macroeconomic policy response and strong global trade institutions, the worst fears have been avoided, at least for the time being (Eichengreen and Irwin 2009).
onetheless, monitoring by Global Trade Alert (Evenett 2010) and the WTO indicates that new protectionist measures continue to be implemented more quickly than in the pre-crisis period – and more quickly than they are removed. How substantially these new protectionist measures are distorting trade has been largely unclear, although the monitoring exercises provide a critical first step in that direction.
Kee et al. (2010) and Bown (2010) provide thorough assessments of the impacts of tariff and trade defence measures, respectively. Here we attempt a more comprehensive empirical assessment of the protectionist response to the crisis1, which also includes export policies, “buy national” policies, bailouts, and other “murky” measures – what Evenett termed the “diversity in contemporary protectionism”. This diversity is not amenable to modelling in the partial or general equilibrium frameworks familiar to trade economists. Taking an econometric approach, however, and with the benefit of detailed trade data, we find a strong association between the imposition of a new measure and variations in trade. This association holds even after separating the impact of macroeconomic factors.
A first peak at the data
Our monthly bilateral 4-digit HS merchandise trade data, provided by Global Trade Information Services (GTIS), cover 80% of global merchandise trade from July 2007 to early 2010. We match these data with information on discriminatory measures from Global Trade Alert. The data identify implementing country, affected trade partners, 4-digit HS product category, month of implementation (and removal), and type of measure. We use only “red” measures – those that are clearly discriminatory and have been implemented. We analyse 314 such measures, which we classified as import restrictions, behind-the-border measures, export restrictions, and export support measures. Incomplete data forces us to drop 194 (mostly “behind-the-border”) measures.
We first consider how trade evolved in a particular 4-digit product category following the implementation of a new import restriction. To separate the impact of the new restriction from common product-specific factors within any 4-digit product, we track trade in the affected country-pair combinations relative to global trade in the same product. We normalise to 100 the “market share” of affected country-pairs in the month prior to the implementation of a new import restriction. We expect this “market share” of protected country-pairs to fall in the following months.
Figure 1 organises some of our raw data by the month in which a new restriction was implemented. It illustrates, despite volatility in the detailed trade data, that the imposition of new import restrictions was followed by a persistent decline in the value of trade covered by restrictions beyond that in global trade of the same product
The first step in our formal analysis consists of obtaining econometric results for trade impacts of protectionism at the product level. The econometric analysis allows us to more thoroughly disentangle the impact of the new discriminatory measures from that of other factors. Our estimation method uses time-varying product fixed effects to ensure that our results exploit only the variation between protected and non-protected country pairs, while holding time and product constant. Further specifications then additionally control for importer-specific demand shocks, exporter-specific supply shocks, and changes in any bilateral trade determinants, such as transport connections or the exchange rate.
According to our coefficient estimates for these different fixed-effect specifications, trade subject to new import restrictions and behind-the-border measures declines by (at least) 5% to 9%2.
In a second step, we use our product-level estimates to calculate a headline number for measures’ impact on aggregate global trade. To this end, we multiply product-level coefficient estimates by the value of trade covered by each measure. This calculation implies that measures included in our analysis distorted global trade by about 0.25%, or $35 billion a year. Border import measures and “behind-the-border” measures (such as subsidies and bailouts) each account for about half of our total estimated trade distortion, with a typical behind-the-border measure distorting seven times more trade than a typical border measure.
Our estimate of a 0.25% protectionist distortion confirms that protectionism played only a minor role in the Great Trade Collapse at end-2008. However, the decline is considerable when seen as a recurring distortion or when compared to trade liberalisation initiatives. For instance, other authors’ estimates of the trade impact of draft Doha Round provisions on industrial tariffs and agriculture centre around 1% of global goods trade.
Putting bounds on “headline” numbers
We are handicapped by missing information on one-third of “red” measures implemented during our study period. Our estimate of a total 0.25% trade distortion by new measures is likely biased downward by excluding them. While we cannot know the extent of this bias, we can provide upper bounds for the total protectionist impact using back-of-the-envelope arithmetic.
If the excluded measures are distributed similarly to the measures included in our analysis and have effects similar to those of the included measures, our estimate would increase to 0.35%, or $50 billion a year. The excluded measures are, however, disproportionately behind-the-border measures. This is important because a single typical behind-the-border measure distorts more trade than does a typical border import measure. We cannot provide a point estimate, but if all excluded measures are behind-the-border and the same product-level coefficients apply, our estimate of the total annual impact would be 0.75% of world trade ($110 billion).
Dissecting the estimates
Digging more deeply into the baseline estimates, we separate results by policy type and by (imposing or affected) region.
Policy types. Behind-the-border bailout and subsidy measures are estimated to account for about half of the aggregate distortion. Tariffs and import bans are responsible for about one fourth, with other border measures such as anti-dumping and public procurement accounting for the rest.
Geography. Whether imposed by advanced or developing countries, new measures are strongly associated with a decline in observed trade. Advanced countries’ measures account for two thirds of the aggregate trade distortion, with nearly all the rest attributable to those of upper middle-income countries. Likewise, advanced and developing countries both had their exports adversely affected by the measures of others. We estimate that developing countries suffer two fifths of the export loss, with this share split evenly between upper-middle and lower-middle income countries (and thus little impact on low-income countries). Advanced countries suffer the remaining three-fifths of the total export loss.
Our aggregate results suggest that new trade measures implemented through early 2010 have not interfered substantially with the global recovery. Nevertheless, two aspects of our findings should raise cautionary flags.
- There is strong empirical evidence that new measures, and in particular unconventional measures, are distorting product-level trade.
As long as they remain in place, these measures will impose recurring costs.
- Individual measures have been potent, typically reducing affected trade flows by 5% to 9%.
If measures continue to be implemented more quickly than they are removed, they will be a drag on the global recovery.
Policymakers must remain alert. Ongoing currency tensions illustrate that protectionist pressures are not abating and may intensify.
- Advanced country policymakers need to be especially vigilant given high domestic unemployment (Gregory et al. 2010).
- Further enhancing monitoring efforts and increased high-level political awareness will bolster resistance to these pressures.
- Concluding the Doha Round would help greatly by reducing tariff ceilings and by strengthening trade rules in many other areas. It is the surest way to avoid the adverse macroeconomic consequences of a widespread resort to protectionism.
The views expressed herein are those of the authors and should not be attributed to the International Monetary Fund, its Executive Board, or its management.
Bown, Chad P, 2010, “Taking Stock of Antidumping, Safeguards, and Countervailing Duties, 1990-2009”, World Bank Policy Research Working Paper 5436.
Eichengreen, B and D Irwin (2009), “The Slide to Protectionism in the Great Depression: Who Succumbed and Why?” NBER Working Paper 15142.
Evenett, Simon (ed.) (2010), Managed Exports and the Recovery of World Trade: The 7th GTA Report, available from www.globaltradealert.org.
Gregory, Rob, Christian Henn, Brad McDonald, and Mika Saito (2010), “Trade and the Crisis: Protect or Recover”, IMF Staff Position Note 10/07.
Henn, Christian, and Brad McDonald (forthcoming), “Protectionist Responses to the Crisis: Damage Observed in Product-level Trade”, IMF Working Paper.
Kee, Hiau Looi, Cristina Neagu, Alessandro Nicita (2010), “Is Protectionism on the Rise? Assessing National Trade Policies during the Crisis of 2008”, World Bank Policy Research Working Paper 5274.
Shingal, Anirudh (2009), “The Impact of Cross-Border Discrimination on Japanese Exports: A Sectoral Analysis,” in Simon Evenett (ed.), The Unrelenting Pressure of Protectionism: the Third GTA Report, available from www.globaltradealert.org.
1 While comprehensive, our approach relies on dummy variables to identify whether an observation is affected by a new measure and does not incorporate information on the intensity of a measure where that information exists (such as for tariffs). Shingal (2009) also analyses many types of measures in a single framework, but focuses exclusively on twelve Japanese export sectors and uses much more aggregate trade data.
2 Actual declines could be higher to the extent that measures affect very specific products. Our analysis cannot quantify impacts at these very disaggregate levels, which would in all likelihood be higher. This is because GTA codes measures at the (relatively aggregate) HS 4-digit level, conditioning our choice of trade data. Coefficient estimates in our main regressions have the expected negative signs and are highly statistically significant. The only exceptions were export measures, for which estimates were inconclusive and we thus omitted them in the calculation of aggregate impacts. Our forthcoming IMF working paper describes data, results, and methods in more detail.